Description
NTDB / TQP Trauma Registry Study
A risk-adjusted outcomes study built on the National Trauma Data Bank and its successor, the American College of Surgeons Trauma Quality Programs participant use file. Trauma registry data are unusually easy to analyse badly: the injury severity variables are derived rather than measured, the physiology fields are missing in patterns that correlate with the outcome, and the sample is not population-based no matter how large it is. This service is built around those three facts.
Read This First: Data Access
You must hold the ACS data use agreement and supply the data. We do not hold a TQP participant use file agreement and cannot obtain the dataset on your behalf. The application goes through your institution to the American College of Surgeons, and the agreement restricts who may analyse the data and how it may be shared, which means the file has to come from you, under your agreement, with your institution’s approval for our involvement as analysts.
If you have not yet applied, apply first. If you are unsure whether your existing agreement covers external analytic support, check with your institutional contact before purchasing. We would rather you find out now than after payment.
When This Is the Right Study
TQP data are the right instrument for questions about in-hospital trauma outcomes across many centres: mortality and complications by injury pattern, disparities in trauma care, the effect of centre characteristics, resuscitation practice variation, and the epidemiology of specific injury mechanisms at a scale no single centre reaches.
They are the wrong instrument for incidence or population rates. TQP is a convenience sample of participating trauma centres, not a population-based registry, the denominator is unknown, participating hospitals change year to year, and patients who died in the field or were treated at non-participating centres are simply absent. Any paper reporting a national rate from TQP is wrong, and reviewers who work in trauma will say so. There is also no post-discharge follow-up: the record ends at discharge, so long-term function, readmission and mortality after discharge are all unavailable. For the general problem of what public injury data can and cannot support, see our analysis of whether NEISS injury statistics are reliable.
How the Analysis Is Actually Done
- Injury severity derived, not trusted – We map ICD-10-CM diagnosis codes to Abbreviated Injury Scale severities and recompute ISS and NISS ourselves rather than accepting the registry field, which is inconsistently populated across centres and years. The mapping version used is stated, because different versions give different scores for the same patient.
- Physiologic missingness treated as informative – Emergency department GCS, systolic pressure and respiratory rate are missing far more often in patients who arrived intubated, sedated or in extremis. Complete-case analysis therefore drops the sickest patients and biases mortality downward. We examine the missingness pattern, use multiple imputation with the outcome in the imputation model, and report a complete-case sensitivity analysis alongside.
- Facility clustering modelled – Patients within a trauma centre are not independent. We fit mixed-effects models with a facility random intercept, or use generalised estimating equations with robust standard errors, and report the intraclass correlation. Pooled logistic regression that ignores the centre is the most common serious error in this literature.
- Transfers handled by an explicit rule – Transferred-in patients can be double-counted, and their outcomes reflect two institutions. We define the transfer rule up front and test whether it changes the conclusion.
- Risk adjustment on the accepted variable set – Age, mechanism, ISS or NISS, GCS motor component, arrival physiology and comorbidity burden, with observed-to-expected mortality ratios where benchmarking is the question.
- Dead-on-arrival and early-death handling stated – Whether patients dying within the first minutes are included changes results substantially. The rule is stated and its effect tested, rather than chosen quietly.
- Centre level and verification status accounted for – Level I, II and III centres treat different case mixes; verification status is a confounder for anything involving outcome quality.
- Year effects examined – The participating hospital set changes over time, so apparent trends can be composition effects. We test for this before reporting any trend.
What You Receive
- Complete manuscript – Written for your target trauma or surgery journal
- Publication-quality tables and figures – Including the risk-adjusted forest plot and centre-variation caterpillar plot where relevant
- Full analysis code – R or Stata, commented and runnable, including the ICD-to-AIS mapping step and the cleaning pipeline. Related reading: six techniques for cleaning clinical registry data
- Completed STROBE checklist
- Analysis log – Case definition, every exclusion, every derived variable, imputation model specification
- One round of author-requested revisions
What Is Not Included
We do not obtain the data. We do not ghost-write authorship, you set the question, review the interpretation, and own the claims. We do not submit for you, do not guarantee acceptance, and do not handle peer-review response within this package. We cannot analyse identifiable patient data; the participant use file is de-identified and that is what we work with.
When We Will Tell You Not to Buy This
If your question needs population incidence, TQP cannot answer it. If it needs outcomes after discharge, TQP cannot answer it. If your injury subgroup has fewer than a few hundred cases with the outcome of interest, facility-clustered risk adjustment will not converge sensibly. And if the ACS agreement in your hands does not permit external analytic support, we cannot take the work at all. All four are checked before we start.
Timeline
Standard delivery is 15 business days from receipt of the data and an agreed analysis plan. The clock starts when the file arrives, not at purchase, TQP files are large and institutional transfer can take time.
Common Questions
Which years can you use? Whichever your agreement covers. We will advise on whether pooling across the NTDB-to-TQP transition is defensible for your specific variables.
Can you compute a national injury rate? No, and neither can anyone else from this dataset. See the limitation above.
Can you benchmark my centre against the national sample? Yes, with observed-to-expected mortality ratios and appropriate shrinkage, this is one of the things TQP data are genuinely good for.
My submission was rejected for inadequate risk adjustment. Can you rework it? This is a common request and usually costs less than a new study. Send the manuscript and the reviewer comments.
Extended Scope
Multi-year pooled analyses with centre-level covariates, machine-learning mortality prediction with external validation, and linkage to other injury data sources are quoted individually, typically from $3,500. See the Extended & Multi-Database Registry Study.
Have a Question First?
Ask before you buy, particularly about data access. If you are not certain your ACS agreement covers external analytic support, or not sure whether TQP can answer your question at all, send it below. A senior analyst replies within one business day. There is no obligation, and if the honest answer is that this is not the right dataset for you, we will say so.
